[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86013-en":3,"doc-seo-86013-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86013,1099514067415,"Rowan","https://ap-avatar.wpscdn.com/avatar/100002539d78ffe74a7?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779092875211072502",8,"Research & Report","Scaffold Splits Hide Structural-Frontier Failures in ADMET Models","Molecular property models for ADMET are commonly tested by holding out Bemis–Murcko scaffolds, but scaffold identity captures only one notion of chemical unfamiliarity. A label-free structural-frontier split is proposed to reserve the sparsest, most physicochemically remote scaffold groups and evaluate six public ADMET tasks. Compared with a scaffold control using identical acyclic grouping, frontier splits raise equally weighted primary error (taskwise median 87.0%) and skew-sensitive mean 130.3%. Removing BBB reverses ranking for that endpoint, and additional controls suggest training-penalty choices do not resolve the observed frontier failures.","arXiv :2607 . 10729v 1 [ cs .LG] 12 Jul 2026  \nScaffold splits hide structural-frontier failures in ADMET models  \nJiacheng Zheng1,5,*∗, Chang Guo2 , Zixuan Wang3 , Xinyu Liu4 ,  \n1 Ma Yinchu School of Economics, Tianjin University  \n2 Department of Mathematics, Faculty of Mathematical & Physical Sciences, University College London  \n3 School of Life Science and Technology, Harbin Institute of Technology  \n4 School of International Trade and Economics, University of International Business and Economics  \n5 Marine College, Shandong University  \n[karcenzheng@gmail.com](karcenzheng@gmail.com), [zcah072@ucl.ac.uk](zcah072@ucl.ac.uk)  \n[w472878575@gmail.com](w472878575@gmail.com), [xinyu_liu2026@163.com](xinyu_liu2026@163.com),  \n* Corresponding author  \nAbstract  \nMolecular property models are commonly evaluated by holding out Bemis–Murcko scaffolds, yet a scaffold identifier is only one notion of chemical unfamiliarity. We introduce a label-free structural-frontier split that reserves the sparsest and most physicochemically remote scaffold groups, and evaluate it on six public experimental or curated ADMET tasks. Against a 70/10/20 scaffold control with identical acyclic grouping, the frontier inflates equally weighted primary error with a taskwise median of 87.0% and a skew-sensitive mean of 130.3%(descriptive task/seed bootstrap interval, 52.1–246.0%) . The mean falls to 75.9% once BBB is removed; that endpoint is the one whose score ranking inverts at the frontier.  \nA message-passing graph-network control still shows a large gap (mean 82.8% over four tasks) and does not invert, so a low-capacity head does not explain the effect. We also test Multi-View Frontier Risk Extrapolation (MV-FREX), a count-adjusted tail-risk penalty over four molecular views, and treat it as a falsifiable probe. It changes normalized frontier error by only 0.16% relative to empirical risk minimization for the perceptron head (interval, −0 .43–0.84%) and by −1 .9% for the graph network;  \nthree fixed robust-penalty controls are likewise inconclusive. Against the published Lo-Hi and DataSAIL splitters the frontier inflates error more on average, though no split is uniformly hardest. An audit of 31,561 marine natural products further shows that OOD status and agreement with legacy ADMET predictions depend on the molecular view, endpoint and teacher coverage. Split construction and label provenance are important evaluation constraints in their own right, and the tested training penalties do not resolve the frontier failures we observe.  \nKeywords: ADMET; molecular property prediction; out-of-distribution generalization; data splitting; applicability domain; distributionally robust learning; benchmark evaluation.  \nAvailability: Code, split manifests, per-run predictions and figure-generation scripts are provided in the anonymized supplementary archive; a persistent public repository will be released after review.  \n∗Jiacheng Zheng received his bachelor’s degree from Marine College, Shandong University, and is currently pursuing his master’s degree at Ma Yinchu School of Economics, Tianjin University.  \n1 Introduction  \nMachine-learning models for absorption, distribution, metabolism, excretion and toxicity (ADMET) are used to prioritize compounds before expensive assays. Prospective compounds, however, can occupy chemical regions that are sparsely represented in historical data. A random test partition therefore gives an optimistic estimate when close analogues cross folds. The widely used Bemis–Murcko scaffold split mitigates this leakage by grouping molecules around a common ring-and-linker framework [Bemis and Murcko, 1996], but it converts a continuous and representation-dependent support question into a single binary partition.  \nRecent evaluations show that molecular ranking can change substantially with the dataset, representation and split [Deng et al., 2023]; simulated prospective splits [Landrum et al., 2023], low-similarity benchmarks [Steshin","cbCaigG0TcAHnruz","https://ap.wps.com/l/cbCaigG0TcAHnruz","pdf",746191,2,1,25,"English","en",105,"# Abstract\n# Introduction\n## Scaffold splitting and leakage concerns\n## Structural frontier reliability question\n## Label provenance and dataset constraints\n# Availability","[{\"question\":\"Why can Bemis–Murcko scaffold splits misrepresent real-world generalization in ADMET?\",\"answer\":\"Because they turn a continuous, representation-dependent notion of chemical support into a single binary holdout, which can miss failures near remote regions that are both sparsely represented and locally unsupported.\"},{\"question\":\"What is the structural-frontier split introduced in the study?\",\"answer\":\"It is a label-free split that reserves the sparsest scaffold groups that are most physicochemically remote, evaluating model reliability at the structural frontier rather than only on unseen scaffold identities.\"},{\"question\":\"How do the researchers test whether the effect is due to low-capacity modeling or training penalties?\",\"answer\":\"They compare against a message-passing graph-network control, which still shows a large gap and does not invert, and they evaluate MV-FREX plus several robust-penalty controls, which do not resolve the frontier failures observed.\"}]",1784207786,63,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"scaffold-splits-hide-structural-frontier-failures-in-admet-models","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/scaffold-splits-hide-structural-frontier-failures-in-admet-models/86013/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-25","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why can Bemis–Murcko scaffold splits misrepresent real-world generalization in ADMET?","Question",{"text":75,"@type":76},"Because they turn a continuous, representation-dependent notion of chemical support into a single binary holdout, which can miss failures near remote regions that are both sparsely represented and locally unsupported.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the structural-frontier split introduced in the study?",{"text":80,"@type":76},"It is a label-free split that reserves the sparsest scaffold groups that are most physicochemically remote, evaluating model reliability at the structural frontier rather than only on unseen scaffold identities.",{"name":82,"@type":73,"acceptedAnswer":83},"How do the researchers test whether the effect is due to low-capacity modeling or training penalties?",{"text":84,"@type":76},"They compare against a message-passing graph-network control, which still shows a large gap and does not invert, and they evaluate MV-FREX plus several robust-penalty controls, which do not resolve the frontier failures observed.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]